[comp.dcom.sys.cisco] Static Subnet Defaults..

cja@flamingo.nersc.gov (Cathy Aronson) (10/02/90)

I was wondering if anyone else besides us has been having trouble
using static subnet defaults with versions 8.1(14) and/or 8.1(19)?
It is a very strange problem.  Everything works fine until the
box reloads.  Then the network is known via "static" instead of
just the subnet being known via "static" and the network itself
being known via "connected".
When this happens the network itself is no longer distributed
into the appropriate routing processes (even if you try using 
redistribute static).  Our set up worked just fine til
8.1.  

Here is the kicker.  The only way to make the problem go away,
is to configure the box from memory.  I have tried everything else
and conf from mem is the only thing that makes the network known
again via "connected".   I can reproduce this problem over and over
whenever I want.  There is no difference between what is in the
volitile mem, and what is in non-volitile memory at the time 
that I config it from memory.  

The problem happens on two of the three routers that we are using
static default subnet on.  

If anyone has any ideas, or if anyone else has seen this,
please let me know.

Thanks a lot!


	Cathy Aronson
        National Energy Research Supercomputer Center
        Lawrence Livermore National Laboratory
        cja@marmot.nersc.gov
	415-422-4016

fortinp@bwdls56.bnr.ca (Pierre Fortin) (10/03/90)

In article <27385@boulder.Colorado.EDU>, cja@flamingo.nersc.gov (Cathy Aronson) writes:

[stuff deleted]

>                          I can reproduce this problem over and over
> whenever I want.  There is no difference between what is in the
> volitile mem, and what is in non-volitile memory at the time 
> that I config it from memory.  
> 
We've also run across a problem related to non-volatile memory (my coleague
should have reported it directly to cisco by now).  At any rate, my 
conclusion was that the NV memory actually contained lots of stuff which 
is *not* displayed with "show config".
 
To cisco (avoided using caps... :^) ):  Can you provide an optional 
parameter to allow showing *all* contents of the NV memory?
Something like "show config all"; and why not "write term all" while you're
at it?

In our case, a "write erase" quickly followed by "write memory" was required
to get rid of a persistent "no ip route-cache" statement which would 
disappear with "no ip address" and reappear when an IP address was coded.

> 
> 	Cathy Aronson
>         National Energy Research Supercomputer Center
>         Lawrence Livermore National Laboratory
>         cja@marmot.nersc.gov
> 	415-422-4016

Cheers,
Pierre
Pierre Fortin       Bell-Northern Research     I know, my postings are
Internet Systems    P.O.Box 3511, Stn C        terse and humourless. So?
(613)763-2598       Ottawa, Ontario            RIP: aptly named protocol
                    Canada    K1Y 4H7          AppleTalk: Adam&Eve's design

lougheed@cisco.com (10/04/90)

When you do a "show configuration" you do indeed see the entire contents of
NVRAM, except for a "magic number" indicating the memory was written, a byte
count, and a checksum.  We store the configuration commands in NVRAM as
ASCII text.

To avoid consuming too much memory, only the non-default configuration
settings are entered into the NVRAM when you do a "write" command.

In the case you mention, I suspect entering the command "ip route-cache"
would have done the trick.

Kirk Lougheed
cisco


>> We've also run across a problem related to non-volatile memory (my coleague
>> should have reported it directly to cisco by now).  At any rate, my 
>> conclusion was that the NV memory actually contained lots of stuff which 
>> is *not* displayed with "show config".
>>  
>> To cisco (avoided using caps... :^) ):  Can you provide an optional 
>> parameter to allow showing *all* contents of the NV memory?
>> Something like "show config all"; and why not "write term all" while you're
>> at it?
>> 
>> In our case, a "write erase" quickly followed by "write memory" was required
>> to get rid of a persistent "no ip route-cache" statement which would 
>> disappear with "no ip address" and reappear when an IP address was coded.